How to Win Big

[Marc Wachowitz]

While the idea to build (more or less) everything around Lisp has a
certain appeal, this will also very likely inhabit an even smaller
niche than Lisp already inhabits today. For some people around here,
it may be sufficient to have their own somewhat isolated Lisp machine,
but from the feedback, I guess the primary interest of most is having
a decent Lisp environment - both for development and for delivery -
which can covers even very low-level stuff (but doesn't necessarily do
so). It doesn't seem like the existing divergence of interests is likely
to disappear easily any time soon, but the nature of the task (either
way) calls for a cooperation of as many good people as we can get. Yet,
we may be able to have our cake and eat it too.

My suggestion is to consider Richard P. Gabriel's paper on "How to Win
Big" (see e.g. http://www.naggum.no/worse-is-better.html), particularly
the third chapter. At a first glance it might sound as even contrary to
the idea of a "Lisp OS" after which the mailing list was named, but on
closer view, I think it's a realistic approach to get both: A reasonable
chance to be able to use Lisp for commercial purposes more often than
now, and a chance to implement as much of a system as one wants - and as
different a system as one wants - in Lisp.

For both, I think we need a Common Lisp (probably with some additional
support for other Lisp dialects [like continuations for Scheme], or even
quite different languages) with a very flexible interface to the lower
levels of the system (where one variant is that the lower levels will be
mostly written in Lisp itself, making the distinction between "levels" a
matter of conventions), and some high-level stuff which isn't standardized
yet. If we're going to either hack some (existing or even newly created)
Lisp implementation to provide this, we might as well do so with some care
to make interdependencies a little bit more explicit than good engineering
practice would demand anyway. This would probably include a (less formal)
reanimation of the appearently sleeping efforts to extend the range of
what's reasonably standard across existing implementations (travel back
into the past of the newsgroup comp.std.lisp).

-- Marc Wachowitz <mw@ipx2.rz.uni-mannheim.de>